How to Connect Windsurf IDE to Browser MCP for Agent Browsing

Why Connect Windsurf Cascade to a Cloud Browser MCP?

While Cascade can write and edit code in your repository, it has no native window into dynamic web content. Standard HTTP fetch tools fail on modern React/Vue SPAs and get blocked instantly by Cloudflare, DataDome, or Akamai firewalls.

Connecting Windsurf to 2Captcha Browser MCP unlocks:

• Dynamic JavaScript Execution: Cascade inspects rendered DOM trees, client-rendered tables, and interactive elements.
• Automated CAPTCHA & WAF Solving: Turnstile and reCAPTCHA challenges are resolved automatically in the cloud before content reaches Cascade.
• Zero Local Chrome Overhead: Browser instances run on high-performance cloud nodes with clean residential IP routing, keeping your developer machine cool and quiet.

Step 1: Locate or Create Windsurf's MCP Config

Windsurf manages external MCP servers through a single configuration file named mcp_config.json located in your user profile:

You can also open this directly from Windsurf by pressing Cmd/Ctrl + Shift + P and selecting Preferences: Open User Settings (JSON) or navigating to Windsurf Settings > Model Context Protocol.

Step 2: Add 2Captcha Browser MCP Server Entry

Add the twocaptcha-browser server definition inside the mcpServers object. If mcp_config.json already has other tools (e.g. Postgres, GitHub), simply append this block:

{
  "mcpServers": {
    "twocaptcha-browser": {
      "command": "npx",
      "args": [
        "-y",
        "@2captcha/browser-mcp"
      ],
      "env": {
        "TWOCAPTCHA_API_KEY": "your_2captcha_api_key_here"
      }
    }
  }
}

Replace your_2captcha_api_key_here with your active 2Captcha API key from your dashboard.

Step 3: Refresh Cascade & Verify MCP Tools

Save mcp_config.json and restart Windsurf, or press Cmd/Ctrl + Shift + P and run Developer: Reload Window.

Open the Cascade panel. You will now see active browser automation tools registered under MCP:

• browser_navigate: Instructs the cloud browser to open any URL with automated WAF solving.
• browser_snapshot: Captures the rendered DOM tree, text nodes, and computed styles.
• browser_click: Clicks buttons, dropdowns, and triggers client-side events.
• browser_screenshot: Returns full-page PNG captures for UI verification.

Step 4: Example Cascade Prompts

In Cascade agent mode, you can now execute end-to-end browser workflows:

#1 Recommended Browser MCP Provider